EEG is better left alone

Abstract

Automated preprocessing methods are critically needed to process the large publicly-available EEG databases, but the optimal approach remains unknown because we lack data quality metrics to compare them. Here, we designed a simple yet robust EEG data quality metric assessing the percentage of significant channels between two experimental conditions within a 100 ms post-stimulus time range. Because of volume conduction in EEG, given no noise, most brain-evoked related potentials (ERP) should be visible on every single channel. Using three publicly available collections of EEG data, we showed that, with the exceptions of high-pass filtering and bad channel interpolation, automated data corrections had no effect on or significantly decreased the percentage of significant channels. Referencing and advanced baseline removal methods were significantly detrimental to performance. Rejecting bad data segments or trials could not compensate for the loss in statistical power. Automated Independent Component Analysis rejection of eyes and muscles failed to increase performance reliably. We compared optimized pipelines for preprocessing EEG data maximizing ERP significance using the leading open-source EEG software: EEGLAB, FieldTrip, MNE, and Brainstorm. Only one pipeline performed significantly better than high-pass filtering the data.

Pseudo‐mutual gazing enhances interbrain synchrony during remote joint attention tasking

“…Leveraging these computed probabilities, we discarded non‐brain components with a likelihood exceeding 90% of being categorized as muscle, eye, heart, line noise, channel noise, or other sources. This 90% threshold aligns with that utilized in Delorme's “Optimized Pipelines” (Delorme, 2023) and other studies (Bierwirth et al, 2023). The remaining independent components were then back‐projected to complete the signal reconstruction.…”

“…Being distanced from the primary sources of eye movement artifacts, these sites offered a more appropriate choice for the current study. Furthermore, to manage artifact rejection, we utilized independent component analysis (ICA, infomax algorithm) (Makeig et al, 1995) (Delorme, 2023) and other studies (Bierwirth et al, 2023). The remaining independent components were then backprojected to complete the signal reconstruction.…”

Evaluating the effectiveness of artifact correction and rejection in event-related potential research

“…The present study addresses both of these issues, examining whether artifact correction and rejection help to minimize consistent but artifactual differences between conditions (i.e., reduce potential confounds) and help to minimize uncontrolled variance (i.e., increase statistical power). Note that although some prior studies of the effectiveness of artifact correction have assessed uncontrolled variance or statistical power (e.g., Delorme (2023); Klug & Gramann (2021); Mennes et al (2010), previous research has largely neglected the possibility that artifacts are a confound. It is unlikely that any artifact correction method will be perfect, and it is essential to assess whether any residual artifactual signals create meaningful confounds in a given study.…”

“…This method could be utilized in future research to assess the effectiveness of alternative artifact minimization approaches and could be applied to other experimental paradigms and participant populations. Moreover, we sought to test the proposal that artifact correction and rejection may do more harm than good (Delorme, 2023).…”

“…Eyeblinks generate large artifacts in electroencephalographic (EEG) recordings, typically exceeding 200 µV at the frontal pole (Fp1 and Fp2) and 30 µV at the vertex (Cz; see (Lins et al, 1993) for A careful assessment of this combined correction-rejection approach is particularly relevant at this moment because the usefulness of extensive EEG preprocessing has been questioned (Delorme, 2023) and because a new metric of data quality is now available that is directly related to effect sizes and statistical power (Luck et al, 2021). The rejection of trials containing artifacts may be particularly problematic because it will decrease the number of trials included in the averages, making the averaged ERP waveforms noisier.…”